Corr DAGs

Relationships come from this google sheet

source(here::here("code", "update-dag.R"))

dag <- read_csv(here::here("data", "dag.csv"))

dag %>% select(-paper) %>% kablebox()

concept	from	edge	to	cites	cites_empirical
experience	office organization	increases	capacity	NA	NA
experience	professional networks	increases	capacity	NA	NA
power	capacity	increases	level of constituency service	NA	NA
experience	experience in office	increases	office organization	NA	NA
experience	experience in office	increases	professional networks	NA	NA
experience	experience in office	increases	policy expertise	NA	NA
demand	experience in office	increases	name recognition	NA	NA
demand	capacity	increases	advertising services	NA	NA
demand	advertising services	increases	demand	NA	NA
power	bureaucratic responsiveness	increases	marginal benefit of service work	NA	NA
power	marginal benefit of service work	increases	directing resources to service	NA	NA
power	directing resources to service	increases	capacity	NA	NA
experience	policy expertise	increases	priority for policy work	NA	NA
power	institutional position	increases	bureaucratic responsiveness	NA	NA
power	institutional position	increases	power over policy	NA	NA
power	power over policy	increases	marginal benefit of policy work	NA	NA
power	marginal benefit of policy work	increases	priority for policy work	NA	NA
power	institutional position	increases	staff	NA	NA
power	staff	increases	capacity	NA	NA
power	priority for policy work	increases	ratio of policy to service	NA	NA
demand	power over policy	increases	media coverage	NA	NA
demand	media coverage	increases	name recognition	NA	NA
demand	name recognition	increases	demand	NA	NA
demand	demand	increases	level of constituency service	NA	NA
demand	advertising services	increases	name recognition	NA	NA

More & less complex pathways

library(netlit)

set.seed(13)

# plot tidy network data
plot_netework <- function(dag){
  d <- review(dag,
              edge_attributes = names(dag) )

n <- ggnetwork(d$graph)

n$name %<>% str_replace_all("([a-z| ]{8}) ", 
                           "\\1\n") %>% str_to_title()



ggplot(n) +
  aes(x = x, y = y, xend = xend, yend = yend,
      label = name ) +
    geom_nodes(size = 15, alpha = .1) +
    geom_edges(aes(#linetype = concept,
                   color = concept), 
               curvature = 0.1, 
               alpha = .8,
               #box.padding = unit(1, "lines"),
             arrow = arrow(length = unit(6, "pt"), type = "closed")) +
  
    geom_nodetext_repel(size = 2.5) +
  theme_blank() + 
  labs(color = "",
       linetype = "") + 
  scale_color_viridis_d(option = "plasma", begin = 0, end = .9, direction = -1) + 
  theme(legend.position="bottom")
}

plot_netework(dag)

drop_node <- function(data, NODE_TO_DROP){
simple <- data %>% mutate(
  from = case_when(
    from == NODE_TO_DROP ~ dag$from[which(dag$to == NODE_TO_DROP)],
    T ~ from
  ),
  to = case_when(
    to == NODE_TO_DROP ~ dag$from[which(dag$to == NODE_TO_DROP)],
    T ~ to
  )
)
return(simple)
}

simple <- drop_node(dag, "directing resources to service") %>%
  drop_node(NODE_TO_DROP = "marginal benefit of policy work") %>% 
  drop_node("bureaucratic responsiveness") %>% 
  drop_node("media coverage") %>% 
  drop_node("policy expertise")

set.seed(7)

#  plot_netework(simple)


# now dropping core mechanisms that don't confound
simpler <- drop_node(simple, "office organization") %>%
  drop_node("professional networks") %>% 
  drop_node("marginal benefit of service work") %>% 
  drop_node("ratio of policy to service") %>% 
drop_node("bureaucratic responsiveness") %>% 
  drop_node("staff")

set.seed(2)
plot_netework(simpler)

simpler %>% 
    filter(!str_detect(to, "advert|name"),
           !str_detect(from, "advert|name") ) %>%
    # drop_node("power") %>%
  plot_netework()

simpler %>% 
    filter(!str_detect(to, "demand|advert|name"),
           !str_detect(from, "demand|advert|name") ) %>%
  plot_netework()

Level Model

priority_vars <- "priority|ratio|policy "

level_only <- . %>% 
  filter(! str_detect(to, priority_vars))


set.seed(30)

dag %>% 
  level_only() %>%
  drop_node("power over policy") %>% 
  plot_netework()

# simple %>% 
#   level_only() %>%
#   drop_node("power over policy") %>% 
#   plot_netework()

#set.seed(49)

set.seed(55)


simpler %>% 
  level_only() %>%
  drop_node("power over policy") %>% 
  plot_netework()

Ratio Model

level_vars <- "level|responsiveness|service |resources to service|capacity|demand|advertising|professional|name|organization|media|staff"

priority_only <- . %>% 
  filter(! str_detect(to, level_vars))

dag %>% 
  priority_only() %>%
  #drop_node("power over policy") %>% 
  plot_netework()+ scale_color_viridis_d(option = "plasma", begin = 0, end = .5, direction = -1)

# simple %>% 
#   priority_only() %>%
#   #drop_node("power over policy") %>% 
#   plot_netework()+ scale_color_viridis_d(option = "plasma", begin = 0, end = .5, direction = -1)

simpler %>% 
  priority_only() %>%
  #drop_node("power over policy") %>% 
  plot_netework()+ scale_color_viridis_d(option = "plasma", begin = 0, end = .5, direction = -1)